About that curious plateau and the long BAT OS

[biopearl123]

A possible explanation for the prolonged post-2019 Kuykendall/Moffitt survival plateau. I tried to see of there was a consistant approach to patients leaving registry for another study that could explain the plateau and tick marks by looking at earlier Moffett studies. Here is Chat GPT. MG maybe if you agree see of Claude agrees.

A key methodological question in interpreting the post-2019 Kuykendall/Moffitt myelofibrosis survival curve is how patients who later entered investigational studies—particularly IMpact-MF—were handled.

In the earlier Moffitt analysis, Between a Rux and a Hard Place (2018), the cohort was explicitly defined as patients who received and discontinued ruxolitinib outside the context of a clinical trial, establishing a “clean” post-rux natural-history cohort at baseline. Importantly, this exclusion applied only to initial rux exposure and does not clarify how patients were handled if they later entered an investigational study after rux failure.

Methodologically, it would be improper to exclude such patients at baseline, because at the moment of rux failure it is impossible to know prospectively which patients will later survive long enough, stabilize sufficiently, and qualify for a clinical trial such as IMpact-MF. Excluding them retroactively would introduce selection bias.

The proper natural-history approach is therefore to include all eligible patients at rux failure and then follow them forward. Patients who die remain events. Patients who later leave standard care for transplant, outside care, or investigational study enrollment are typically censored at the point of departure.

This distinction is important.

If patients in the post-2019 cohort were followed from rux cessation and then censored when entering IMpact-MF, their survival time up to trial entry would remain in the Kaplan-Meier curve, but their subsequent outcomes—including death—would not. Those later deaths would not be retroactively added back into the original analysis.

This provides a plausible explanation for the unusual shape of the post-2019 curve: an early steep drop (patients too ill to bridge) followed by a prolonged plateau with multiple tick marks (patients surviving long enough to transition into investigational studies).

If correct, the plateau may not represent pure “BAT-like” natural-history survival, but rather survival truncated at the point of clinical trial entry. This would tend to lengthen the apparent post-rux overall survival and complicate direct comparison with randomized event-driven studies such as IMpact-MF.

[biopearl123]

the key sentence from the ASCO coverage of that abstract is:

“earlier switch to next-line JAKi or clinical trial enrollment”